arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2502.01459stat.MEcs.LGstat.ML
收藏

学习对序列进行部分延迟决策

Learning to Partially Defer for Sequences

Department of Statistics University of Michigan(统计系密歇根大学)

Sahana Rayan, Ambuj Tewari

AI总结:

针对序列预测,提出两种基于模型的事后拒绝器,实现细粒度部分延迟决策,在多个任务上优于整体延迟,改善成本-准确性权衡。

摘要:

在“学习延迟决策”(L2D)框架中,预测模型要么做出预测,要么根据一个拒绝器(rejector)的决定将预测推迟给专家。当前的L2D方法训练拒绝器来决定是否拒绝{\em整个预测},这在模型预测长序列时并不理想。我们提出了一种针对序列输出的L2D设置,在该设置中,系统可以将整个模型预测的\textit{特定输出}推迟给专家,以便在预测过程中将专家和机器交错进行。我们为预训练预测器提出了两种基于模型的事后拒绝器:一种是令牌级拒绝器(token-level rejector),它将特定的令牌预测推迟给具有下一令牌预测能力的专家;另一种是一次性拒绝器(one-time rejector),适用于不具备此类能力的专家,它将从特定点开始的剩余序列推迟。在实验中,我们还实证表明,在旅行商问题求解器、新闻摘要和天气预报任务上,这种细粒度的延迟决策比整体延迟决策能实现更好的成本-准确性权衡。

英文摘要:

In the Learning to Defer (L2D) framework, a prediction model can either make a prediction or defer it to an expert, as determined by a rejector. Current L2D methods train the rejector to decide whether to reject the {\em entire prediction}, which is not desirable when the model predicts long sequences. We present an L2D setting for sequence outputs where the system can defer \textit{specific outputs} of the whole model prediction to an expert in an effort to interleave the expert and machine throughout the prediction. We propose two types of model-based post-hoc rejectors for pre-trained predictors: a token-level rejector, which defers specific token predictions to experts with next token prediction capabilities, and a one-time rejector for experts without such abilities, which defers the remaining sequence from a specific point onward. In the experiments, we also empirically demonstrate that such granular deferrals achieve better cost-accuracy tradeoffs than whole deferrals on Traveling salesman solvers, News summarization, and Weather prediction.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑